A Framework for Quantifying and Reducing Uncertainty in InfoSymbiotic Systems Arising in Atmospheric Environments
Abstract
This project focuses on large scale dynamic data driven applications systems (DDDAS, or InfoSymbiotic systems) governed by partial differential equations (PDEs), e.g., arising in atmospheric environments. Specifically, our main interests are data assimilation and the configuration of sensor networks. During this project we have developed a rigorous framework for quantifying and reducing uncertainty in the context of InfoSymbiotic systems. This includes a goal-oriented aposteriori error estimation methodology for the impact of different errors on the variational solutions of inverse problems; an optimization-constrained optimization problem approach to find the optimal configuration of the DDDAS system; new parallel-in-time algorithms to speed up variational inference; a trust-region approach to perform inference in an ensemble space; new nonlinear filtering and smoothing algorithms that sample directly from the posterior PDF using a Hybrid Markov-Chain Monte Carlo(HMCMC) approach; and a solid theoretical basis for optimization with reduced order models.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 29, 2016
- Accession Number
- AD1004746
Entities
People
- Adrian Sandu
Organizations
- Virginia Tech